The Autonomous Finance Project
An independent research programme into autonomy, governance and systemic consequences.
Explore published research questions, curated institutional evidence, study references and controlled teaching simulations.
From an institutional research question to an experimental record and a testable decision boundary. Different instruments. One discipline: publish the assumptions, preserve the limits, and make the work reproducible.
Four research instruments, each with a separate question: autonomous finance governance, matched counterfactual experiments, agent authority and the reconciliation boundary between traditional and tokenised accounting records.
An independent research programme into autonomy, governance and systemic consequences.
Explore published research questions, curated institutional evidence, study references and controlled teaching simulations.
Test an intervention against a common baseline under matched assumptions.
Run deterministic paired trials in execution, collateral and signal triage. Change the seed, test sensitivity and export fully specified research packets.
Turn abstract decision authority into rules that can be challenged.
Define a financial mandate, evaluate synthetic requests, examine deny-overrides and human-review rules, and run an adversarial boundary suite.
Two accounting systems. One economic claim. Reconciliation is the experiment.
Inspect reserve locking, dual acknowledgement, issuance, transfer, redemption, delayed finalisation and deliberately faulty adapters. Trace every synthetic state, test the invariant boundary and export the full evidence record.
A controlled simulation can show what follows from an explicit model; it cannot establish what happened at a real bank. A rule evaluation can show whether a synthetic instruction satisfies stated limits; it cannot grant production authority. Each instrument exports its inputs, model version, limitations and outcome so that interpretation remains separable from computation. Read the research registry.